Quantum
Machine Learning Conference 2025
Agenda (times in UTC+1)
11:00: "Introduction to Quantum Machine Learning", Paweł Gora
11:30: "Measuring and Interpreting the States of Temporal Quantum Models", Jacob Cybulski
12:00: "Data preparation for Quantum Autoencoders", Sebastian Zając
12:30: "KANQAS: The Implications of Kolmogorov-Arnold Networks in Quantum Domain", Akash Kundu
13:00: "The Cost of Quantum Kernel Methods: Insights into Resource Estimation", Artur Miroszewski
13:30: "Exploring Quantum Machine Learning and Quantum Federated Learning for Financial Fraud Detection", Nouhaila Innan
14:00: "Automated Quantum Machine Learning as Hyperparameter Optimization Problem", Tomasz Rybotycki
14:30: "Towards Quantum Neural Network implementation in semiconductor single-electron devices", Krzysztof Pomorski
Speakers
Paweł Gora is the Founder and CEO of the Quantum AI Foundation aiming to support education, research, and collaboration in new technologies (especially quantum computing and AI). He is also the Chairperson of the Board of QWorld. More info: https://www.mimuw.edu.pl/~pawelg.
"Introduction to Quantum Machine Learning"
Abstract: In this talk, I will give an introduction to quantum machine learning. First, I will explain what quantum computing is and how it is different from classical computing paradigm. Later, I will tell how machine learning could be enhanced using quantum computing, presenting some of the possible opportunities and existing challenges. Finally, I will present how machine learning could be applied in the quantum computing domain, and outline how one can continue education and start a professional career in the quantum machine learning domain.
Jacob Cybulski is the founder of Enquanted, providing research, training and consulting services in quantum computing. He is also Honorary Associate Professor in Quantum Computing, School of IT Deakin University, Australia. Jacob’s projects involve quantum machine learning for understanding complex data, as well as, for prediction and anomaly detection in business, engineering, and science applications. He is dedicated to education and promotion of quantum technology. Jacob’s work also concerns classical machine learning and data visualisation. In the past, Jacob also conducted research at the University of Melbourne, La Trobe University and RMIT University, Australia.
"Measuring and Interpreting the States of Temporal Quantum Models"
Abstract: Jacob’s recent presentations focused on the analysis of time-series and signals using quantum machine learning. They covered considerations of quantum models design, methods of embedding classical data, training of quantum models, and the analysis of temporal data. This talk, however, discusses some practical and theoretical issues of measuring the state of a temporal quantum model, and the subsequent interpretation of measurement results in the context of various analytic tasks.
Sebastian Zając is an assistant professor at the Decision Analysis and Support Unit of the SGH Warsaw School of Economics in Poland. He is a theoretical physicist and researcher keen to explore the intersection of cutting-edge technologies and practical business applications. His research focuses on AI methodology in credit risk, MLops practices in a cloud environment, Quantum Machine Learning, and Quantum Computing on NISQ devices. His dedication lies in exploring the potential of quantum computing to solve complex problems with time series and graph data.
"Data preparation for Quantum Autoencoders"
Abstract: This presentation will focus on the development and application of quantum machine learning methods, with particular emphasis on encoding strategies for time-series data. We will explore the key challenges and solutions related to data encoding techniques, including how to optimally represent time-series data within a quantum model. The talk will cover various architectural choices for quantum autoencoders. Additionally, practical aspects of model training, such as selecting optimizers, loss functions, and methods for executing quantum models on simulators, GPU accelerators, and QPUs, will be discussed. We will also highlight the importance of testing and validating these models to ensure effective performance in real-world scenarios.
Akash Kundu is a prominent quantum researcher currently serving as a postdoctoral researcher at the University of Helsinki. His work primarily focuses on quantum architecture search and the application of reinforcement learning to enhance variational quantum algorithms. Kundu has contributed significantly to the field, with multiple publications addressing optimizing quantum circuits and their performance in noisy intermediate-scale quantum (NISQ) environments. Kundu's academic background includes a Bachelor of Physics from the University of Calcutta, during which he developed a keen interest in quantum physics. He has participated in various research projects that bridge machine learning with quantum computing, demonstrating his versatility and commitment to advancing technology in these areas during his PhD.
In addition to his research, Kundu has been involved in teaching and mentoring roles, helping to foster the next generation of researchers in artificial intelligence and quantum technologies.
"KANQAS: The Implications of Kolmogorov-Arnold Networks in Quantum Domain"
Abstract: The design of efficient quantum circuits is a pivotal challenge in quantum computing. KANQAS introduces a novel framework that enhances quantum architecture search (QAS) by integrating Kolmogorov-Arnold Networks (KANs) into deep Q-networks, replacing traditional Multi-Layer Perceptrons (MLPs). This talk will outline the KANQAS methodology and its application to tasks such as quantum state preparation and quantum chemistry problems. Notably, KANQAS demonstrates a 2–5 times higher probability of success than MLPs in noiseless scenarios, requiring fewer learnable parameters and exhibiting greater robustness against noise. We will explore the implications of KANQAS for achieving quantum advantage and discuss future research directions. By presenting this innovative approach, we aim to highlight its potential to significantly advance the field of quantum circuit design and its applications in real-world quantum computing.
A. Miroszewski received a Ph.D. degree in theoretical physics from the National Centre for Nuclear Research, Otwock, Poland, in 2021. He is a Postdoctoral Researcher with the Jagiellonian Uni-
versity, Krakow, Poland. He is involved in European Space Agency projects exploring the potential of quantum machine learning for satellite data analysis and serves as a quantum computing lecturer
at the IEEE GRSS HDCRS summer schools. His main research interests include the application of kernel methods for classification tasks.
"The Cost of Quantum Kernel Methods: Insights into Resource Estimation"
Abstract: Quantum kernel methods have emerged as promising tools in machine learning, particularly for tasks where computing kernel matrices efficiently can lead to significant advantages. These methods utilize quantum kernel estimation, which is theoretically believed to provide computational speedups, including exponential gains, in the construction of kernel matrices. Kernel matrices, essential in both supervised and unsupervised learning algorithms such as kernelized principal-component analysis and support vector machines (SVMs), reinforce many machine learning workflows.
In this talk, I will shortly review my work on the performance of quantum kernel SVMs application to remote sensing. Then, I will highlight the crucial role of resource estimation in evaluating the feasibility of quantum kernel methods, focusing particularly on the determination of the number of shots required in quantum circuits to achieve reliable kernel estimations. A key challenge here is the presence of barren plateaus and the phenomenon of exponential value concentration, which can severely impact the efficiency of quantum computations.
To address these challenges, I will introduce methods to estimate the number of shots necessary for accurate quantum kernel computation. I will present results from resource estimation analyses applied to a typical classification problem, demonstrating the practical implications for quantum machine learning. Finally, I will draw conclusions on the scalability and potential of quantum kernel methods, offering insights into their role in advancing machine learning applications.
[1] Havlíček, Vojtěch, et al. "Supervised learning with quantum-enhanced feature spaces." Nature 567.7747 (2019): 209-212.
[2] A. Miroszewski et al., Quantum machine learning for analyzing multi- and hyperspectral satellite images (2024). (https://quantumcosmos.org/wp-content/uploads/2024/11/postdoc_report_final.pdf)
[3] Thanasilp, Supanut, et al. "Exponential concentration in quantum kernel methods." Nature Communications 15.1 (2024): 5200.
[4] Miroszewski, Artur, et al. "In Search of Quantum Advantage: Estimating the Number of Shots in Quantum Kernel Methods." arXiv preprint arXiv:2407.15776 (2024).
Nouhaila Innan is a Research Team Lead at eBRAIN Lab and a Postdoctoral Associate at the Center for Quantum and Topological Systems (CQTS) at New York University (NYU) Abu Dhabi. She earned her PhD in Quantum Machine Learning from Hassan II University of Casablanca, where she also completed her Bachelor's in Physics & Applications and Master's in Physics & New Technologies, specializing in materials and nanomaterials. Her research focuses on quantum machine learning, quantum algorithms, and their applications in fields like finance and cybersecurity. Innan is passionate about mentoring and making quantum technologies accessible through global initiatives.
"Exploring Quantum Machine Learning and Quantum Federated
Learning for Financial Fraud Detection"
Abstract: As financial fraud becomes increasingly complex and widespread, the demand for advanced detection methods to address these evolving challenges is more critical than ever. Quantum Machine Learning (QML) introduces innovative approaches to model development, offering significant potential for enhancing fraud detection systems. This talk addresses how QML frameworks can tackle the challenges of non-standardized and complex financial datasets. It also introduces a Quantum Federated Learning architecture designed to enhance security and enable collaborative model development while preserving privacy. By analyzing performance metrics and configurations on real-world datasets, this discussion highlights QML's potential to improve detection accuracy, scalability, and privacy in fraud detection while outlining key areas for future exploration.
BIO: Dr. Tomasz Rybotycki received his doctoral degree in 2023 from the Systems Research Institute of the Polish Academy of Sciences in Warsaw. His supervisor was Professor Piotr Kulczycki, PhD.
During his PhD, Dr. Rybotycki worked on predictive density estimation of non-stationary data streams, where he designed an algorithm based on a kernel density estimator for predictive data density analysis. The title of his dissertation was “Data density estimation for non-stationary data streams.” During his doctoral studies, he also earned a bachelor's degree in physics, after successfully defending his undergraduate thesis entitled “Framework for performing experiments on IBM Quantum Computers.”
Since 2017, Dr. Rybotycki has been working at the Systems Research Institute of the Polish Academy of Sciences, currently as an Assistant Professor. He also works at CAMK PAN as a specialist-technician and at CDSI AGH as an AI specialist (research specialist).
In 2020, Dr. Rybotycki was awarded a research fellowship at AstroCeNT, where he researched the application of metaheuristics for learning quantum neural networks. In the same year, he joined the Center for Theoretical Physics of the Polish Academy of Sciences, first as a researcher/software engineer and then as a PhD student (theoretical physics) in the project. He left the Center for Theoretical Physics in 2022.
In 2022, Dr. Rybotycki joined ACK Cyfronet AGH, where, as a senior programmer, he was responsible for preparing an e-platform for automated quantum machine learning (AQMLator).
Dr. Rybotycki rejoined AstroCeNT (and later CAMK) in 2023. He is currently working on the development of quantum machine learning techniques with Piotr Gawron, PhD. His research focuses on optimizing quantum circuits using ZX Calculus and applying (Q)ML techniques to Earth observations.
He is currently completing his master's degree in philosophy at the UW Department of Philosophy, where his main focus is on philosophical aspects of science (particularly quantum mechanics and quantum computing) and logic. As part of her master's thesis, she is investigating the philosophical implications of inverse, backward and negative time (physical concepts) in relation to the ontology of time.
Automated Quantum Machine Learning as Hyperparameter Optimization Problem
Abstract: A successful Machine Learning (ML) model implementation requires three main components: training dataset, suitable model architecture and training procedure. Same goes for Quantum ML (QML) models, although here the problem is slightly more specific. Often, when looking for an architecture, we're actually looking for a Parametrized Quantum Circuit (PQC) structure, or --- in the context of Quantum Neural Networks (QNN) --- we talk about Quantum Architecture Search (QAS) problem. In this talk, I'll present AQMLator, an AutoQML platform of our design, and present QAS as a Hyperparameter Optimization (HPO) problem. I'll also mention possible applications of such an approach.
BIO: Krzysztof Pomorski received the Ph.D. degree from the Faculty of Physics, Astronomy and Applied
Computer Science, Jagiellonian University, Krakow, Poland, in 2015 and 2 Master of Science degrees in Physics from University of Lodz and Electronics from Lodz University of Technology. During his Ph.D. research, he developed the scheme of field-induced Josephson junctions and pointed out the existence of topological Meissner effect. He also developed microscopic justification of canonical quantization of Josephson effect. He is specialized in fundamental modeling of devices with the use of physics methodology and in the development of new numerical algorithms. He held a postdoctoral position working on design of superconducting RAM for fluxon electronics at Nagoya University and at AGH University of Science and Technology and at the same time, he was continuing the cryogenic experiments.
He modeled semiconductor singe-electron devices for quantum chips at University College Dublin, Cracow University of Technology, Lodz University of Technology and currently at Quantum Hardware Systems company. He proposed the implementation of interface between quantum semiconductors and quantum superconducting computer. He has founded the company Quantum Hardware Systems in Lodz and YouTube channel of the same name. He was giving lectures on Quantum Technologies, Artificial Intelligence and Artificial Life. He was supervisor of 3 Master of Science thesis, 2 Engineering and 11 Bachelor degrees and organizer of 5 conferences. Currently he is involved in the design of superconducting hybrid classical-quantum computer and in study of quantum neural networks. The interlink between classical and quantum information theory, many-body physics, quantum field theory, statistical physics is expecting to bring unique synergy between theoretical physics, and the development of new quantum technologies.
Towards Quantum Neural Network implementation in semiconductor single-electron devices
Abstract: Review of existing approach towards quantum neural networks is given [1-3]. Philsophy of implementation of charged-based single-electron quantum artificial neural networks using neighbouring semiconductor quantum dots controlled electrostatically is given [4]. Logical gates represented by various electrostatical potential configurations are being modelled numerically in classical and in quantum regime.
References:
[1]. Kerstin Beer, Quantum neural networks, Arxiv: 2205.08154, defended PhD Thesis. 2022.
[2]. R. W. Newcomb, C. H. Yang and A. M. Hodge, "Quantum dot neural network neurons," 7th International Conference on Control, Automation, Robotics and Vision, 2002. ICARCV 2002., Singapore, 2002, pp. 29-33 vol.1, doi: 10.1109/ICARCV.2002.1234785.
[3]. M.V. Altaisky, N.N.Zolnikova, N.E. Kaputkina, V.A. Krylov, Y.E.Lozovik, N.S.Dattani
Towards a feasible implementation of quantum neural networks using quantum dots
Appl. Phys. Lett. 108, 103108, 2016.
[4]. K.Pomorski, Electrostatically Interacting Wannier Qubits in Curved Space, Materials 2024, 17.
Organizers:
